2.1 Ultrafast solid-state lasers - ETH - the Keller Group
2.1 Ultrafast solid-state lasers - ETH - the Keller Group
2.1 Ultrafast solid-state lasers - ETH - the Keller Group
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Ref. p. 134] <strong>2.1</strong> <strong>Ultrafast</strong> <strong>solid</strong>-<strong>state</strong> <strong>lasers</strong> 101<br />
[99Mat]. The strong periodic variation in <strong>the</strong> group delay of <strong>the</strong> original chirped mirrors occurs<br />
due to impedance mismatch between <strong>the</strong> incident medium (i.e. typically air) and <strong>the</strong> mirror stack<br />
and also within <strong>the</strong> mirror stack. Using <strong>the</strong> accurate analytical expressions for phase, group delay,<br />
and GDD [99Mat], DCMs could be designed and fabricated with a smooth and custom-tailored<br />
GDD suitable for generating pulses in <strong>the</strong> two-cycle regime directly from a Ti:sapphire laser [99Sut,<br />
99Mor1, 99Mor2].<br />
A Double-Chirped Mirror (DCM) [97Kae, 99Mat] is a multilayer interference coating that can<br />
be considered as a composition of at least two sections, each with a different task. The layer<br />
materials are typically SiO 2 and TiO 2 . The first section is <strong>the</strong> AR coating, typically composed of<br />
10 to 14 layers. It is necessary because <strong>the</strong> <strong>the</strong>ory is derived assuming an ideal matching to air.<br />
The o<strong>the</strong>r section represents <strong>the</strong> actual DCM structure, as derived from <strong>the</strong>ory. The double-chirp<br />
section is responsible for <strong>the</strong> elimination of <strong>the</strong> oscillations in <strong>the</strong> GDD from within <strong>the</strong> mirror stack.<br />
Double-chirping means that in addition to <strong>the</strong> local Bragg wavelength λ B <strong>the</strong> local coupling of <strong>the</strong><br />
incident wave to <strong>the</strong> reflected wave is independently chirped as well. The local coupling is adjusted<br />
by slowly increasing <strong>the</strong> high-index layer thickness in every pair so that <strong>the</strong> total optical thickness<br />
remains λ B /2. This corresponds to an adiabatic matching of <strong>the</strong> impedance. The AR-coating,<br />
toge<strong>the</strong>r with <strong>the</strong> rest of <strong>the</strong> mirror, is used as a starting design for a numerical optimization<br />
program. Since <strong>the</strong>oretical design is close to <strong>the</strong> desired design goal, a local optimization using<br />
a standard gradient algorithm is sufficient. At this point, only <strong>the</strong> broadband AR-coating sets a<br />
limitation on <strong>the</strong> reduction of <strong>the</strong> GDD oscillations. An AR-coating with a residual reflectivity of<br />
less than 10 −4 is required for a DCM at a center wavelength of around 800 nm, which results in<br />
a bandwidth of only 250 nm [00Mat]. This bandwidth limitation cannot be removed with more<br />
layers in <strong>the</strong> mirror structure [96Dob].<br />
The invention of <strong>the</strong> BAck-SIde-Coated (BASIC) mirrors [00Mat] or later <strong>the</strong> tilted front-side<br />
mirrors [01Tem] resolved this issue. In <strong>the</strong> BASIC mirror <strong>the</strong> ideal DCM structure is matched to<br />
<strong>the</strong> low-index material of <strong>the</strong> mirror which ideally matches <strong>the</strong> mirror substrate material. This<br />
DCM structure is deposited on <strong>the</strong> back of <strong>the</strong> substrate and <strong>the</strong> AR-coating is deposited on <strong>the</strong><br />
front of <strong>the</strong> slightly wedged or curved substrate, so that <strong>the</strong> residual reflection is directed out of <strong>the</strong><br />
beam and does not deteriorate <strong>the</strong> dispersion properties of <strong>the</strong> DCM structure on <strong>the</strong> o<strong>the</strong>r side<br />
of <strong>the</strong> substrate. Thus, <strong>the</strong> purpose of <strong>the</strong> AR-coating is only to reduce <strong>the</strong> insertion losses of <strong>the</strong><br />
mirror at <strong>the</strong> air–substrate interface. For most applications it is sufficient to get this losses as low<br />
as 0.5 %. Therefore, <strong>the</strong> bandwidth of such an AR-coating can be much broader. Both <strong>the</strong> DCM<br />
and AR-coating multilayer structures can be independently designed and optimized. Based on this<br />
analysis, we designed a BAck-SIde-Coated Double-Chirped Mirror (BASIC DCM) that supports<br />
a bandwidth of 220 THz with group delay dispersion oscillations of about 2 fs 2 (rms), an order-ofmagnitude<br />
improvement compared to previous designs of similar bandwidth [00Mat]. Ultrabroad<br />
BASIC DCMs with a bandwidth of 270 THz have also been used to compress supercontinuum of<br />
cascaded hollow fibers down to 4.6 fs [04San]. The trade-off is that <strong>the</strong> substrate has to be as thin<br />
as possible to minimize <strong>the</strong> overall material dispersion. In addition, <strong>the</strong> wedged mirror leads to an<br />
undesired angular dispersion of <strong>the</strong> beam.<br />
Ano<strong>the</strong>r possibility to overcome <strong>the</strong> AR-coating problem is given with <strong>the</strong> idea to use an ideal<br />
DCM under Brewster-angle incidence [03Ste]. In this case, <strong>the</strong> low-index layer is matched to air.<br />
However, under p-polarized incidence <strong>the</strong> index contrast and <strong>the</strong>refore <strong>the</strong> Fresnel reflectivity of a<br />
layer pair is reduced and more layer pairs are necessary to achieve high reflectivity. This increases<br />
<strong>the</strong> penetration depth into <strong>the</strong> mirror which has <strong>the</strong> advantage that <strong>the</strong>se mirrors can produce<br />
more dispersion per reflection but this means that scattering and o<strong>the</strong>r losses and also fabrication<br />
tolerances become even more severe. In addition, this concept is difficult to apply to curved mirrors.<br />
Fur<strong>the</strong>rmore, <strong>the</strong> spatial chirp of <strong>the</strong> reflected beam has to be removed by back reflection or an<br />
additional reflection from ano<strong>the</strong>r Brewster-angle mirror.<br />
O<strong>the</strong>r methods to overcome <strong>the</strong> AR-coating problem are based on using different chirped mirrors<br />
with slightly shifted GTI oscillations that partially cancel each o<strong>the</strong>r. Normally, <strong>the</strong>se chirped<br />
mirrors are very difficult to fabricate [00Mat]. Many different growth runs normally result in strong<br />
Landolt-Börnstein<br />
New Series VIII/1B1